Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

news.ycombinator.com

Post reply on HN
Print view View original

21–30 of 93 posts

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#21

Post by brap »

endisneigh wrote: post #13

Imho this is a good product, but wouldn’t it make more sense to simply sign videos cryptographically? Unfakeable and unbeatable. So someone uploads a video, you sign it, they display video. If authenticity is in question check the signature. Deep fake detection is intractable imo. Use cryptography instead. Hell if you want to be thorough sign each frame and create an extension for YouTube and other providers to liter…

Most videos people watch nowadays don't come from "official" authorities, whose public keys are known and signatures can be verified. When videos are recorded by random people from their smartphones, what signature are you going to verify?

Even if smartphone manufacturers start integrating digital signatures right into their cameras, you can use the smartphone to re-record a pre-recorded fake video. And I'm sure that with enough resources you can do something way more clever.

I really don't see how crypto can solve this problem. I don't think AI can either (for reasons already mentioned in this thread). It's something we'll have to learn to live with.

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#22

Post by cphoover »

bpcrd wrote: post #19
endisneigh wrote: post #13

Imho this is a good product, but wouldn’t it make more sense to simply sign videos cryptographically? Unfakeable and unbeatable. So someone uploads a video, you sign it, they display video. If authenticity is in question check the signature. Deep fake detection is intractable imo. Use cryptography instead. Hell if you want to be thorough sign each frame and create an extension for YouTube and other providers to liter…

We’re fascinated by the potential applications of crypto in content provenance. In this example, a UGC video platform would need a way to initially determine the content hasn’t been manipulated before it’s signed, right? What about a live scenario where a deepfake mimicking an exec calls a manager to wire $10M ( https://www.forbes.com/sites/thomasbrewster/2021/10/14/huge-... ) We totally recognize deepfake detection…

> "We’re fascinated by the potential applications of crypto in content provenance. In"

@bpcrd what's the advantage of using a block-chain to store video fingerprints, to determine provenance, over say a highly performance-optimized immutable centralized or federated system?

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#23

Post by endisneigh »

brap wrote: post #21
endisneigh wrote: post #13

Imho this is a good product, but wouldn’t it make more sense to simply sign videos cryptographically? Unfakeable and unbeatable. So someone uploads a video, you sign it, they display video. If authenticity is in question check the signature. Deep fake detection is intractable imo. Use cryptography instead. Hell if you want to be thorough sign each frame and create an extension for YouTube and other providers to liter…

Most videos people watch nowadays don't come from "official" authorities, whose public keys are known and signatures can be verified. When videos are recorded by random people from their smartphones, what signature are you going to verify? Even if smartphone manufacturers start integrating digital signatures right into their cameras, you can use the smartphone to re-record a pre-recorded fake video. And I'm sure that…

> Most videos people watch nowadays don't come from "official" authorities

I doubt this is true. Most videos are likely disseminated from a handful of companies.

> Even if smartphone manufacturers start integrating digital signatures right into their cameras, you can use the smartphone to re-record a pre-recorded fake video. And I'm sure that with enough resources you can do something way more clever.

Even if a fake video was re recorded it would have a different signature than the “original”. Problem solved.

Literally cryptography is the only solution

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#25

Post by btown »

candiddevmike wrote: post #15
bpcrd wrote: post #11

Earlier quoted context omitted.

In 2017 deepfakes were pretty crude, today the avg person can’t tell a real face from a deepfake generated on a 5 year old iPhone. We expect the tech to continue moving in this direction. So, similar to anti-virus, we're approaching this problem with an iterative, multi-model solution that can evolve with the threat.

So the folks at the forefront of deep fake technology (i.e. the attackers you're targeting) will slip through your product because it lags behind the state of the art (like AV, which you said is the approach you're following), while innocent folks will be caught by it due to a new kafkaesque version of "prove you're not a bot" since you focus on reducing false negatives. Hopefully I can avoid companies using your pro…

Retrospective antivirus-esque techniques are still useful, though, as not every actor is a state-level actor, and even then, forcing state-level actors to "burn" their state-of-the-art exploits/models because previous exploits/models are detected out-of-the-box, slows down the abuse of those actors.

And realistically, since deepfake detection will inevitably be more expensive than captchas or antivirus scanning, this will be adopted by human-in-the-loop organizations for critical processes where threat scoring or moderation is already being applied.

That said - Reality Defender, please train your system on diverse human data sets, do not release models where ethnicity or gender (including gender identity) are nontrivially correlated with deepfake score, and have processes in place from day 1 to allow users to report suspected patterns of bias. The kafkaesque "prove you're not a bot" scenario envisioned by the parent poster is one thing for holistic human-in-the-loop verification processes, and another thing if it suppresses minority voices and minority access to government services.

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#27

Post by smt88 »

cphoover wrote: post #5

I believe deep-fakes to be a serious threat to functioning democracies around the world. I would love to be on the front-lines fighting against this threat. I have submitted my resume: https://cphoover.github.io/ Have you also considered either offering a browser plugin to display contextual warnings attached to video elements? Or thought of working with browser makers? The web/social media is where a ton of fake med…

I worry about the same thing.

I think "blockchain" (or even just a signed, immutable, public database) is mostly a solution in search of a problem, but I do think it may have an application here.

If you can hash a video when it's recorded and publish the hash with a timestamp that can't be forged, you can at least prove that this video existed at least as long ago as that stamp.

That allows you to invalidate any deepfakes produced on top of that video that have later timestamps.

It's not perfect, but it might be one weapon against this stuff.

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#28

Post by cphoover »

smt88 wrote: post #27
cphoover wrote: post #5

I believe deep-fakes to be a serious threat to functioning democracies around the world. I would love to be on the front-lines fighting against this threat. I have submitted my resume: https://cphoover.github.io/ Have you also considered either offering a browser plugin to display contextual warnings attached to video elements? Or thought of working with browser makers? The web/social media is where a ton of fake med…

I worry about the same thing. I think "blockchain" (or even just a signed, immutable, public database) is mostly a solution in search of a problem, but I do think it may have an application here. If you can hash a video when it's recorded and publish the hash with a timestamp that can't be forged, you can at least prove that this video existed at least as long ago as that stamp. That allows you to invalidate any deep…

You can have an immutable centralized/federated system without using "crypto" though "proof-of-work", there are databases out there that provide immutable storage that don't require this overhead.

Providing api access to a signed immutable database makes sense... but I'm not sure how much sense utilizing existing popular cryptocurrencies would make (e.g. bitcoin, etherium)

Re: Launch HN: Reality Defender (YC W22) – Deepfake Detection Platform

#29

Post by Geee »

We have solved this problem already for text with digital signatures. There might be a way to digitally sign speech in real-time and show the signature on a display device, which is captured on video. It could be just a smartphone app or a separate device. This way every video could be proven authentic, even if they originate from unofficial sources.
Post reply on HN